Apache Druid

&gt; irst of all, is it common to use it? I mean do we really need &gt; 1 replicas? In which scenario it helping?
Yes, if you have task replicas &gt; 1, that means the same partition is being read by two tasks and if one task fails, the other task can be queried still
&gt; another question: can somebody explain how synchronization between tasks happens? is it two that run in parallel and consume data but only one that have interval lock succeedes?
There can only be 1 winner. Whichever task completes or publishes the segment first to deep storage wins.
&gt; What is the termination sequence for the second(“standby” task replica?)
It gets killed

Hi <@U04CGQ1P68Z> thanks for the response
may I ask following: when the replica task gets killed it should be marked as failed, isn’t it? I mean is there some differentiation between replica task killed by “success” scenario of prime task and some real failure scenario?

if you have some code pointers, I’d like to investigate some corner case we have regarding this case.

I’ve reread what I’ve written - I meant “shouldn’t marked as failed”